THE VCORE MOD GUIDE

THE VCORE MOD GUIDE

Updated 05/11/02 by Tim B @ OCWorkbench.com

The original guide by Kevin Cribbs
http://www.ocworkbench.com/hardware/elite/k7s5amod/k7s5amodp1.htm

Variable Resistor Guide by Ezra
http://www.ocworkbench.com/ocwbcgi/ultimatebb.cgi?ubb=get_topic&f=4&t=003356

NOTE: Less resistance = higher vcore!, check the potentiometers variable resistance before turning the power on to ensure its at the right end of the scale.

By Ezra.
So I couldn't get my Duron 750 to run at 1000 MHz. It would POST okay but not load the boot sector from my SCSI drive, even with all of the L7 bridges...well, bridged. The CPU temperature was staying below 105F (41C) and it never really did any computation, so I figured that low Vcore had to be the problem.
I read Kevin Cribbs's wonderful article about various possible modifications to the K7S5A. Having no desire to undertake such a difficult, not to mention risky, series of modifications, I looked for an easier, less permanent way to do the same thing. I also have a friend who last week used a potentiometer to boost Vcore on his Shuttle AK31 board.

I had a good look at this picture from Mr. Cribbs's article:

The chip in the picture is near the CPU and the AMR slot.

I went over to Radio Shack (a US chain of electronics stores) and bought a small 47k-Ohm potentiometer (variable resistor) that is "suitable to be mounted on a printed circuit board". I also got hold of a couple tiny clips with 8" (20 cm) leads. Total cost, including tax, was less than 2 USD.

The object of the game here, for those that don't know, is to use a resistor to bleed off a little current from the voltage regulator so that it will raise Vcore to compensate. Since we want to keep Vcore as low as possible--don't want to fry the Duron!--while still achieving the desired clock speed, we need a way to adjust the resistance. So we use an adjustable resistor (the potentiometer) and twiddle it until we get what we want.

Anyway, I soldered the lead from one clip to the center leg of the potentiometer and the lead from the other clip to one of the other legs (it doesn't matter which). I wrapped the potentiometer in electrician's tape, insulating the solder points but leaving the little dial exposed so I could adjust it.

I turned the dial clockwise as far as it would go and measured the resistance from one clip to the other with my multimeter, getting a reading of 44.0k-Ohm. Since Mr. Cribbs had expressed some concern that using a potentiometer in this application could introduce unknown capacitance or inductance into the system, I also used my multimeter to measure the capacitance of my device. The capacitance was below the minimum that my multimeter could measure, 0.002nf. I assumed that a stray 2pf is too small to have an effect on the K7S5A. I do not have the means to measure inductance, so I resolved to keep the leads nice and straight.

Next I referred to the picture above and clipped one lead to pin 1 and the other to pin 7 of the voltage regulator chip. Noting that the tiny clips didn't grip the pins of the chip very firmly, I used a little electrician's tape to keep everything in place. I then turned on the machine.

Success! Well, it powered on and started to POST, anyway. I hit Del and proceeded to the BOIS hardware monitor to look upon what I had wrought. It was good. Vcore was reading 1.640v, up from 1.600v. Everything else looked as stable as ever.

Since I had already tried pencilling the L7 bridges without success, I knew to dial the potentiometer so that Vcore went above 1.840v, the highest Vcore I had achieved with my pencil. I dialed up 1.880v and tried to boot.

The machine booted fine and I played around for about 15 minutes before it reset itself. Needless to say I hit Del again and went to the hardware monitor.

After making sure that heat was not a problem (CPU temp was below 40C), I dialed the Vcore a little higher. I found that the smallest adjustment I could make was from 1.880v to 1.925v. Either the $0.79 potentiometer has some granularity so that a finer adjustment is not possible, or the K7S5A hardware monitor cannot resolve a smaller difference in Vcore. I booted anyway and gave it a try.

This time I had REAL success! I ran the system under a normal load for an hour or so, noting that (thanks to my FOP38 and a tiny amount of Arctic Silver) CPU temp never exceeded 105F (41C). Then for load testing I played Tribes2 for an hour, noting much better performance and that CPU temp never exceeded 107F (42C). In my opinion a robust FPS game is in many ways a better system test than most benchmark software. The action is heavy and continuous, and all components of the system are used all the time, particularly with an on-line game like Tribes2.

So for about $1.78 and an hour of pleasant work I made my computer 33% faster.

(Many, many thanks to Kevin Cribbs for the article and photo, and to OCWorkBench for hosting it all. I hope no one minds that I swiped the picture.)